← Back to list
Security & Privacy
#SSI#DID#검증가능한 자격증명#선택적 공개#eIDAS
Last updated · 2026-11-09

Self-Sovereign Identity (SSI) and Decentralized Identifiers (DID)

1. Overview

A. Definition

Self-Sovereign Identity (SSI) is an identity model that returns to the individual (the subject)—rather than to an issuing authority or platform—the rights to own, store, and present identity information; the core elements that technically realize it are the Decentralized Identifier (DID) and the Verifiable Credential (VC).

SSI is not a single product but a design philosophy and architectural paradigm. Where conventional identity systems were structured so that "someone identifies and manages me on my behalf," SSI aims at a structure where "I carry my own identity data in a wallet and present it directly when needed." That is, the sovereignty over creation, storage, and control of identity rests with the individual, and the issuer merely attests to a credential and cannot look into how that credential is used.

This shift is not a mere technology swap but a change in how trust is intermediated. Where the past relied on institutional trust—"a trustworthy central authority vouches for it"—SSI moves the basis of assurance to mathematical trust, namely a "cryptographically verifiable signature." Because a credential is trusted not out of faith in the institution but because the signature can be verified, there is no need to query the institution during verification, and the intermediary's involvement and room for tracking shrink accordingly.

This paradigm is often summarized in three design principles. The first is User Control: the individual decides what information to present, to whom, and when. The second is Minimal Disclosure: where only "adult status" is needed, one selectively proves only the fact "19 or older" rather than the full date of birth. The third is Issuer-independent Verification: authenticity is confirmed cryptographically without querying the issuer's server at verification time. Only when these three principles hold simultaneously can something be called SSI; if even one is missing, it remains a centralized service that is SSI in name only.

B. Background and Necessity

Today most digital services operate on a centralized (or federated) identity model. An individual holds accounts scattered across dozens of sites, and each service replicates and stores identifiers such as a resident registration number, email, or phone number in its own DB. In this structure, identity information becomes a primary target for attackers, and in reality a large share of major personal-data breaches aim at these central stores. The more data piles up in one place, the larger the damage upon breach—a "honeypot" problem structurally embedded in the design.

The federated model (social login, etc.) eased account proliferation but bred another dependency. As authentication concentrated on a few giant Identity Providers (IdPs), the user's login behavior is exposed directly to the IdP (which service was accessed, and when), and an IdP outage became a single point of failure that paralyzes logins across many services at once. Convenience was obtained at the cost of mortgaging privacy and availability to giant platforms.

At the root of this dependency lies the structural problem that "the party that uses identity data (the service) and the party that should own it (the individual) are not separated." The individual cannot know where their identity data is replicated or how it is used, and after consent, control effectively disappears. Consent remains as a formality, but there is no real self-determination over "data once handed over." SSI is precisely an attempt to recover this "separation of ownership and use" for identity data; instead of replicating and handing data to services, it inverts the flow itself so that data is kept in a personal wallet and presented only when needed.

On top of this, as stronger data-sovereignty and privacy regulation (the Personal Information Protection Act, GDPR, the "Data 3 Acts," etc.) intertwines with accelerating non-face-to-face transformation, demand for technology to "prove a credential without handing over the original" has exploded. For example, when opening an account by mobile or accessing adult content, if one can present only the verified facts "this is me and I am an adult" instead of uploading the full image of an ID to a server, both breach risk and regulatory burden fall at once. SSI/DID targets exactly this point, drawing attention as a solution that binds the era's demands—data minimization, decentralization, and self-control—into a single identity architecture.

2. SSI/DID Architecture and the Trust Triangle

A. Overall Structure

The SSI ecosystem consists of three parties—Issuer, Holder, and Verifier—and a Verifiable Data Registry (VDR) that serves as their trust anchor. This structure is called the Trust Triangle.

flowchart TD
  I["Issuer(university·government·bank)"] -->|"issue VC(sign)"| H["Holder(personal wallet)"]
  H -->|"present VP(selective disclosure)"| V["Verifier(service provider)"]
  V -->|"trust"| I
  I -.->|"register DID·public key·schema"| R[("VDR(blockchain·distributed ledger)")]
  V -.->|"resolve DID·look up public key"| R
  H -.->|"register·update DID"| R

The core of this structure is that the holder sits at the center of the data flow. In the conventional model, the verifier queried the issuer (or IdP) directly—"Is this the right person?"—and in that process the issuer could know when and where the user used their identity. By contrast, in SSI the issuer issues a VC to the holder once and does not get involved thereafter. The holder presents the VC kept in their wallet directly to the verifier, and the verifier looks up only the issuer's public key from the VDR to verify the signature. Because the issuer cannot know where the credential it issued was used, the privacy-intrusion path of "user tracking by the issuing authority" is blocked at the source.

The VDR need not necessarily be a blockchain or distributed ledger, but thanks to its decentralized, immutable, and highly available properties it is often adopted as the trust anchor. What matters is the design principle that no personal information is put on the VDR. The VDR records only publicly shareable metadata such as DID identifiers, public keys, credential schemas, and revocation status, while the actual credentials (VCs) and personal data are kept entirely in the personal wallet. If this separation is not observed, "personal data is permanently recorded on an un-erasable ledger," colliding head-on with GDPR's right to erasure—so it is the first safeguard to check in SSI design.

B. The VC Issuance–Presentation–Verification Process

In practice, operation proceeds as a sequential flow of issuance, storage, presentation, and verification. Below is the scenario "a university issues a graduation-proof VC, and the individual presents it to a company when applying for a job."

sequenceDiagram
  participant H as Holder Wallet
  participant I as Issuer(university)
  participant R as VDR(distributed ledger)
  participant V as Verifier(company)
  I->>R: register DID·public key·schema
  H->>I: request graduation-proof VC(submit own DID)
  I->>I: verify eligibility then sign VC with private key
  I->>H: deliver signed VC(store in wallet)
  V->>H: request proof of graduation(Presentation Request)
  H->>H: selectively disclose only needed fields(build VP)
  H->>V: present VP(includes holder signature)
  V->>R: resolve issuer·holder DID, look up public keys
  V->>V: verify signature·revocation status·validity period
  V-->>H: verification result(grant access)

In this flow, cryptographic trust is guaranteed by a dual-signature structure. The issuer signs the VC content with its own private key to prove issuer authenticity—"I vouch for this credential"—and the holder, at presentation time, signs the VP (Verifiable Presentation) with their own private key to add a Proof of Possession—"this VC is indeed mine." Because the verifier checks both signatures, it simultaneously prevents a replay attack (stealing and presenting another's VC) and a forgery attack (tampering with the content).

Also, the Presentation Request the verifier sends at the presentation stage specifies "signed by which issuer, from which type of credential, and which fields are needed." The holder's wallet interprets this request, selects a matching VC among those held, obtains the user's explicit consent on "what to present to whom," and only then generates the VP. Thanks to this explicit "request–consent–presentation" flow, the user, at every moment of presentation, sees and approves their own data flow with their own eyes—this is how "self-control" is realized at the UX level.

In particular, selective disclosure at the presentation stage is the highlight of SSI. Even if a graduation-proof VC contains student ID, major, GPA, and graduation date all together, when a company requires only "graduation status," the holder builds and presents a VP containing only the fact "graduated," with the rest hidden. If techniques such as zero-knowledge proofs (ZKP) or BBS+ signatures are used here, one can prove that the issuer's signature over the whole VC is still valid without exposing the remaining fields. This technically completes the minimal-disclosure principle of "only the necessary facts, nothing more."

C. Components

Component Role Key characteristic
DID Decentralized identifier (did:method:identifier) Created without an issuer, self-controllable
DID Document Public keys·service endpoints linked to a DID Looked up by resolving at the VDR
VC Verifiable credential Issuer signature, includes a set of claims
VP Verifiable presentation VC (in part)+holder signature, selective disclosure
Wallet Personal identity wallet (app·secure area) Stores keys·VCs, controls presentation
VDR Verifiable data registry Stores public keys·schemas·revocation info

Rather than each element carrying meaning independently, they interlock to form a single trust system. If the DID is the decentralized identifier of "who," the DID Document carries that DID's public keys to provide "the material to verify the signature," the VC attests "what credential exists" with the issuer's signature, and the VP seals the act "I now present that credential minimally" with the holder's signature. The wallet is the trust-execution point that stores and controls all this personal data and keys, and the VDR is the public trust anchor that no single party monopolizes.

Prying a bit further into the internal structure of a VC makes clear where trust originates. A VC largely consists of three parts: metadata (issuer DID·validity period·type), a set of claims (assertions about the subject), and a proof (the issuer's signature). A claim is an assertion in subject–predicate–object form such as "this subject's degree is a bachelor's," and the proof is the issuer's digital signature over the entire assertion. Therefore the verifier does not trust the claim content itself but trusts the issuer that signed the claim, and whether to trust the issuer is decided by a separate trust framework. Because "data authenticity (signature)" and "issuer authority (governance)" are separated this way, the technical layer can evolve with cryptography and the social-trust layer with governance, each independently. This is the structural foundation that lets SSI expand into an open ecosystem without being locked to a particular platform.

3. Core Technical Elements

A. The DID Identifier and DID Methods

A DID is a URI of the form did:<method>:<method-specific-id>. The method is the DID Method that specifies by which VDR and rules that DID is created and resolved; hundreds exist, such as did:web (web-domain based), did:key (the public key itself as the identifier), did:ion (Bitcoin layer 2), and did:ethr (Ethereum). Because each method differs in degree of decentralization, cost, and performance, choosing a method suited to the purpose—a permissioned-ledger-based method where strong trust is needed, as for a public ID, and a lightweight method such as did:web for lightweight services—is the starting point of design. For example, did:web leverages the trust of an existing web domain (HTTPS·domain ownership) as is, enabling rapid adoption without a blockchain, but it depends on the domain operator and so is not fully decentralized. Conversely, a ledger-based method has strong decentralization and immutability but incurs large registration·resolution cost and latency. Since "decentralization ↔ performance·cost" are thus conflicting axes, the correct order is to first define the required level of trust and then choose a method to match.

The essential value of a DID is that one can create an identifier by oneself, without an issuing authority. An email address is granted and revoked by the mail provider, a phone number by the carrier, but a DID is created together the moment the individual generates a key pair, requiring no one's permission. This shift of identifier control from the issuing authority to the individual is the technical basis for the name "self-sovereign."

Resolving a DID returns a DID Document, which carries the public keys for signature verification (verificationMethod), interaction endpoints (service), references to authentication means, and more. A key design point is that the DID Document contains only public keys, no private keys or personal data. Because a public key is safe even if exposed, it is put on the VDR for anyone to look up, while the private key—the source of signing authority—never leaves the secure area of the personal wallet. This "public key public, private key isolated" principle is the last line of defense against forgery even in a decentralized environment. Moreover, when key rotation is needed, only the DID Document's public key is updated, so one can separate the lifespans of a long-lived identity and short-lived keys while keeping the identifier (DID) itself.

B. Selective Disclosure and Zero-Knowledge Proofs

The selective disclosure seen earlier is implemented by two strands of technology. One is hash-based selective disclosure (SD-JWT, etc.), which hashes each claim and discloses only the original text of the items to be presented. It is simple to implement but cannot do range proofs such as "19 or older" and must disclose the value itself. The other is zero-knowledge proofs (ZKP)·BBS+ signatures, which can prove only the predicate "19 or older" is true without disclosing the date of birth at all. The latter is stronger on privacy but higher in computational cost and implementation difficulty, so one weighs the trade-off according to the service's sensitivity and performance requirements.

The concrete effect can be gauged in numbers. For instance, in resident-ID-based adult verification, around ten fields—name, resident registration number, address, date of birth, and so on—are exposed all at once, but with SSI's range proof one reduces exposed fields to zero and transmits only the true/false 1 bit of "adult status." It amounts to compressing the exposed information from tens of fields down to the 1-bit level, and the damage upon breach falls correspondingly. This design of separating "the target of proof" from "the scope of disclosure" is the decisive reason SSI is rated as privacy-preserving identity beyond a mere digital ID.

C. Revocation and Key Management

A VC may need mid-term revocation even within its validity period (loss of eligibility, loss of the credential, etc.). So that the verifier can confirm revocation without asking the issuer each time, a widely used technique puts a Status List on the VDR·endpoint in bitmap form and has the verifier look up only the relevant index. If the status of hundreds of thousands of credentials is expressed as one compressed bitmap, the verifier can judge revocation immediately by checking just a single bit at a given index, lowering both lookup load and tracking risk. Here too, designing a privacy-preserving status list so that "whose credential was revoked" is not revealed is a challenge.

Meanwhile, since all trust hangs on the private key, recovery upon key loss (social recovery·backup) and use of the device's secure area (Secure Enclave·TEE) become decisive for SSI's practical adoption. In a centralized system the issuer resets things even if you forget your password, but in the self-sovereign model the absence of such a central recovery desk is a double-edged sword. Hence social recovery—distributing recovery shares among trusted acquaintances·institutions—or a function to encrypt and back up·migrate wallet data is required as a virtually mandatory safeguard. Because SSI deployed without a recovery design carries the fatal usability problem of "identity permanently lost the instant it is lost," from a professional-engineer perspective revocation·recovery·key management must be treated with weight equal to the choice of cryptographic algorithm.

4. Comparison with Existing Identity Models

To understand SSI's place, one must set it on the evolutionary axis of identity models. The flow from centralized → federated → self-sovereign is a history in which control has moved from the service, through the platform, to the individual.

Category Centralized Federated (social login) Self-sovereign (SSI)
Identity storage Each service's DB IdP central Personal wallet
Controlling party Service IdP (platform) The individual
Traceability Per service IdP grasps the whole Hard to trace
Single point of failure Service DB IdP Distributed (weak)
Data minimization Difficult Limited Selective disclosure

The differences in the table stem not from a mere feature comparison but from a difference in power structure—"who can track." The reason the IdP sees all of a user's access history in the federated model is that the IdP intervenes as an intermediary at every login. The reason SSI makes tracking hard is the reverse of the same logic. After issuance the issuer drops out of intermediation and the holder presents directly, so no single party stands in a position to amass the entire usage trajectory. That is, the structure guarantees privacy, rather than relying on the goodwill of the operating party.

The difference in data minimization also stems from structure. The centralized and federated models have an incentive to "gather as many attributes as possible" for identification, whereas in SSI selective disclosure—pulling out only the needed attributes at each presentation—is the default, so collection minimization becomes the default. That SSI enforces the "minimum collection necessary for the purpose" required by regulation at the architecture level rather than as an operational rule raises its regulatory fit.

That said, this comparison also reveals SSI's trade-offs. In exchange for a weaker single point of failure, key-management responsibility is shifted onto the individual, and in exchange for control coming to the individual, the recovery·usability burden grows. It is for this reason that real mobile-ID projects cite "re-issuance upon loss·device change" and "usability for the elderly" as the greatest hurdles. Apart from technical superiority, how to rescue an individual who has lost their key is the real bottleneck for SSI going mainstream.

5. Standardization Trends and Practical Adoption

SSI is no longer at the experimental stage but is entering international standards and national identity infrastructure.

On the standards side, the center is the W3C. DID Core finalized the data model for DID identifiers and DID Documents as a standard (Recommendation) in 2022, and the Verifiable Credentials Data Model—after 1.1 was established as a standard—has advanced 2.0 to the Candidate Recommendation stage, codifying the three-party issuer·holder·verifier ecosystem and data minimization. For credential formats, the mdoc of ISO/IEC 18013-5 (mDL, mobile driving licence) and IETF's SD-JWT VC form the two main axes, while for presentation·issuance protocols OpenID4VP/OpenID4VCI has effectively settled as a standard. That said, specific format adoption and versions are changing rapidly, so it is safe to re-confirm the latest specifications at the time of application.

A point to note here is that the SSI ecosystem has not converged on a single standard. For credential formats alone, W3C VC (JSON-LD)·SD-JWT VC·ISO mdoc coexist, each using different signing·disclosure techniques. Thus in practice "which format·protocol combination is adopted" determines the scope of interoperability, and a multi-format wallet that supports several formats together emerges as a realistic solution. That the standards are still evolving is an opportunity for early adoption and, at the same time, a risk that a wrong bet returns as a sunk cost.

On the national-infrastructure side, the most notable case is the EU's eIDAS 2.0 and the EUDI Wallet (European Digital Identity Wallet). eIDAS 2.0 (Regulation (EU) 2024/1183) is a legal framework requiring every member state to provide citizens with a digital identity wallet; it is implemented in stages targeting around 2026 and nudges public and private services to accept this wallet. This is assessed as the world's largest-scale attempt to institutionalize SSI principles into national-level law·infrastructure (specific timelines·detailed rules may be adjusted by implementing acts, so this is described in generalized terms).

Domestically as well, the Ministry of the Interior and Safety's mobile ID (mobile driving licence·national-merit certificate, etc.) is being issued and verified on a DID basis, and DID services and decentralized-identity consortia centered on financial-sector·telecom coalitions continue. A lesson commonly stressed in practical adoption is that the triad of "wallet usability·recovery·offline verification" is as important as technical standards. However cryptographically complete it may be, it fails if an ID cannot be verified in the field without a network or if the elderly find the wallet hard to use.

Industry-specific use cases are also taking concrete shape. In education, attempts continue to issue degrees·certificates as VCs to realize tamper-free career proof and lifelong-learning record management; in finance, a reusable KYC model—perform KYC (customer verification) once, receive it as a VC, and reuse it across institutions—draws attention for cost savings. In healthcare, a form of carrying vaccination·test results in one's own wallet and presenting them across borders, and in supply chain·e-commerce, a form of proving the credentials of firms·products as VCs to screen trustworthy counterparties, are under review. What these cases share is "reusing a once-issued verified credential across many contexts," and this reusability—eliminating the repeated cost of issuance·verification—underpins SSI's economic value.

6. Considerations and Implications

SSI/DID has entered a technically mature stage, but for its value to be realized, technology, institutions, and usability must interlock together. From a professional-engineer perspective, the following should be judged comprehensively when considering adoption.

  • Adoption strategy (hybrid first, transition later): Rather than a full switch to SSI, it is realistic to run in parallel with existing IdP·authentication systems and apply gradually, starting with high-value·high-sensitivity areas (credentials·adult verification·cross-border identity). Nail down a data-minimization design that puts no personal data on the VDR from the outset, to forestall regulatory collisions such as the right to erasure.
  • Trade-off (sovereignty vs. usability): The more control is returned to the individual, the greater the key loss·recovery·device change burden. Finding the balance point of "sovereignty and convenience" through social recovery, secure-area (TEE) backup, and offline-verification design divides business success from failure.
  • Interoperability and governance: If DID methods·VC formats·protocols proliferate, wallets become incompatible and value halves. Standards conformance (W3C·ISO·OpenID, etc.) and the design of a Trust Framework·governance that decides "which issuers to trust" are as important as the technology.
  • Legal effect and liability: In a structure where credentials are presented·verified in a distributed way, "who among issuer·holder·verifier bears liability for damage from a faulty VC" is an area still to be settled. Legal·operational risks—alignment with electronic-signature·electronic-document law, and preparing fallbacks for verification failure—must be examined together with the technical design.
  • Outlook and related technologies: SSI is expected to expand by combining with PETs such as zero-knowledge proofs·homomorphic encryption, blockchain·decentralized infrastructure, biometrics·FIDO/Passkey, and credential delegation for AI agents. In particular, as demand grows for a "verified identity delegable to an agent" in environments where AI agents transact on a person's behalf, SSI is likely to broaden its reach even into machine identity.

References


In one line: SSI/DID is an identity paradigm that returns ownership·control of identity to the individual, realizing—through the issuer·holder·verifier trust triangle and VC·VP·selective disclosure—a data-minimizing, decentralized identity system that "proves only the necessary facts without handing over the original."